4.3 Postdigitization Algorithms for Improving Spurious Free Dynamic Range
As previously mentioned, there are numerous postdigitization techniques for optimizing the
quantization process. Many of these techniques are used to increase the SNR of the quantizer by
improving the predictor characteristics of differential quantization schemes. Another area of
postdigitization processing provides compensation for the nonlinearities that occur in practical
implementations of ADCs. As discussed in Section 2 of this report, these nonlinearities produce
spurious signals that can reduce significantly the SFDR of the ADC. The purpose of this
compensation is to suppress the spurious responses below the noise in the frequency band from 0
to f
s
/2. Two of these techniques, phase-plane and state variable compensation, are discussed
below [13], [14].
Both techniques are used to identify a set of correction factors that can be used to compensate for
any nonlinearity throughout the full amplitude range of the ADC. In phase-plane compensation,
the procedure for correcting the digitized signal is as follows: The input signal is split into two
separate signals. One signal is fed into the ADC and the other is sent through an analog
differentiator and then digitized by a second ADC. The differentiated signal is used to determine
the instantaneous slope of the signal. The output of both ADCs then is used to determine the
correction factor to be applied to the ADC output representing the digitized input signal. A table
consisting of correction factors for each possible combination of quantization level and
instantaneous slope is developed for an individual ADC based on measurements of that
particular ADC. This table then is stored in RAM and is used to provide the correct ADC output
for any given input signal amplitude. Studies using this technique show as much as a 15- to 16-
dB (about 2.5 bits) improvement in the SFDR over uncompensated ADCs [14]. This
improvement, however, is restricted to a narrow frequency band well below f
s
/2.
In an effort to improve the SFDR for all frequencies in the 0 to f
s
/2 frequency band, a state
variable compensation technique was also proposed. This type of compensation is implemented
by applying the input signal to an ADC and splitting the output of the ADC into two signals. One
of these signals is used without modification while the other is delayed by a single clock cycle
(one sample of the input signal). The two outputs, representing the quantization levels for the
present and previous ADC outputs, then are used to determine the correction factor to be applied
to the present ADC output. A table of correction factors for each possible combination of present
and previous quantization levels is developed for an individual ADC based on measurements of
that particular ADC. As in phase-plane compensation, this table is stored in RAM and is used to
provide the correct ADC output for any given input signal amplitude. Tests using this technique
also show as much as a 16-dB improvement in the SFDR over the entire 0 to f
s
/2 frequency band
for the particular sampling rate.
While compensation techniques require additional hardware and testing of individual ADCs,
they can improve the SFDR of the ADC significantly without increasing its resolution (number
of bits). In essence, they bring the characteristics of the ADC closer to the theoretical expectation
37
of its performance. An underlying assumption in these techniques, however, is that the ADC
characteristics are static. Testing of ADCs has shown that for most ADCs this is a valid
assumption [14].
Chia sẻ với bạn bè của bạn: |